2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the AF_INET socket handler.
8 * Version: @(#)sock.h 1.0.4 05/13/93
10 * Authors: Ross Biro, <bir7@leland.Stanford.Edu>
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Florian La Roche <flla@stud.uni-sb.de>
16 * Alan Cox : Volatiles in skbuff pointers. See
17 * skbuff comments. May be overdone,
18 * better to prove they can be removed
20 * Alan Cox : Added a zapped field for tcp to note
21 * a socket is reset and must stay shut up
22 * Alan Cox : New fields for options
23 * Pauline Middelink : identd support
24 * Alan Cox : Eliminate low level recv/recvfrom
25 * David S. Miller : New socket lookup architecture.
26 * Steve Whitehouse: Default routines for sock_ops
27 * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
28 * protinfo be just a void pointer, as the
29 * protocol specific parts were moved to
30 * respective headers and ipv4/v6, etc now
31 * use private slabcaches for its socks
32 * Pedro Hortas : New flags field for socket options
35 * This program is free software; you can redistribute it and/or
36 * modify it under the terms of the GNU General Public License
37 * as published by the Free Software Foundation; either version
38 * 2 of the License, or (at your option) any later version.
43 #include <linux/config.h>
44 #include <linux/list.h>
45 #include <linux/timer.h>
46 #include <linux/cache.h>
47 #include <linux/module.h>
48 #include <linux/netdevice.h>
49 #include <linux/skbuff.h> /* struct sk_buff */
50 #include <linux/security.h>
52 #include <linux/filter.h>
54 #include <asm/atomic.h>
56 #include <net/checksum.h>
59 * This structure really needs to be cleaned up.
60 * Most of it is for TCP, and not used by any of
61 * the other protocols.
64 /* Define this to get the sk->sk_debug debugging facility. */
65 #define SOCK_DEBUGGING
67 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && ((sk)->sk_debug)) \
68 printk(KERN_DEBUG msg); } while (0)
70 #define SOCK_DEBUG(sk, msg...) do { } while (0)
73 /* This is the per-socket lock. The spinlock provides a synchronization
74 * between user contexts and software interrupt processing, whereas the
75 * mini-semaphore synchronizes multiple users amongst themselves.
80 struct sock_iocb *owner;
84 #define sock_lock_init(__sk) \
85 do { spin_lock_init(&((__sk)->sk_lock.slock)); \
86 (__sk)->sk_lock.owner = NULL; \
87 init_waitqueue_head(&((__sk)->sk_lock.wq)); \
93 * struct sock_common - minimal network layer representation of sockets
94 * @skc_family - network address family
95 * @skc_state - Connection state
96 * @skc_reuse - %SO_REUSEADDR setting
97 * @skc_bound_dev_if - bound device index if != 0
98 * @skc_node - main hash linkage for various protocol lookup tables
99 * @skc_bind_node - bind hash linkage for various protocol lookup tables
100 * @skc_refcnt - reference count
102 * This is the minimal network layer representation of sockets, the header
103 * for struct sock and struct tcp_tw_bucket.
106 unsigned short skc_family;
107 volatile unsigned char skc_state;
108 unsigned char skc_reuse;
109 int skc_bound_dev_if;
110 struct hlist_node skc_node;
111 struct hlist_node skc_bind_node;
116 * struct sock - network layer representation of sockets
117 * @__sk_common - shared layout with tcp_tw_bucket
118 * @sk_zapped - ax25 & ipx means !linked
119 * @sk_shutdown - mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
120 * @sk_use_write_queue - wheter to call sk->sk_write_space in sock_wfree
121 * @sk_userlocks - %SO_SNDBUF and %SO_RCVBUF settings
122 * @sk_lock - synchronizer
123 * @sk_rcvbuf - size of receive buffer in bytes
124 * @sk_sleep - sock wait queue
125 * @sk_dst_cache - destination cache
126 * @sk_dst_lock - destination cache lock
127 * @sk_policy - flow policy
128 * @sk_rmem_alloc - receive queue bytes committed
129 * @sk_receive_queue - incoming packets
130 * @sk_wmem_alloc - transmit queue bytes committed
131 * @sk_write_queue - Packet sending queue
132 * @sk_omem_alloc - "o" is "option" or "other"
133 * @sk_wmem_queued - persistent queue size
134 * @sk_forward_alloc - space allocated forward
135 * @sk_allocation - allocation mode
136 * @sk_sndbuf - size of send buffer in bytes
137 * @sk_flags - %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, %SO_OOBINLINE settings
138 * @sk_no_check - %SO_NO_CHECK setting, wether or not checkup packets
139 * @sk_debug - %SO_DEBUG setting
140 * @sk_rcvtstamp - %SO_TIMESTAMP setting
141 * @sk_no_largesend - whether to sent large segments or not
142 * @sk_route_caps - route capabilities (e.g. %NETIF_F_TSO)
143 * @sk_lingertime - %SO_LINGER l_linger setting
144 * @sk_hashent - hash entry in several tables (e.g. tcp_ehash)
145 * @sk_backlog - always used with the per-socket spinlock held
146 * @sk_callback_lock - used with the callbacks in the end of this struct
147 * @sk_error_queue - rarely used
148 * @sk_prot - protocol handlers inside a network family
149 * @sk_err - last error
150 * @sk_err_soft - errors that don't cause failure but are the cause of a persistent failure not just 'timed out'
151 * @sk_ack_backlog - current listen backlog
152 * @sk_max_ack_backlog - listen backlog set in listen()
153 * @sk_priority - %SO_PRIORITY setting
154 * @sk_type - socket type (%SOCK_STREAM, etc)
155 * @sk_localroute - route locally only, %SO_DONTROUTE setting
156 * @sk_protocol - which protocol this socket belongs in this network family
157 * @sk_peercred - %SO_PEERCRED setting
158 * @sk_rcvlowat - %SO_RCVLOWAT setting
159 * @sk_rcvtimeo - %SO_RCVTIMEO setting
160 * @sk_sndtimeo - %SO_SNDTIMEO setting
161 * @sk_filter - socket filtering instructions
162 * @sk_protinfo - private area, net family specific, when not using slab
163 * @sk_slab - the slabcache this instance was allocated from
164 * @sk_timer - sock cleanup timer
165 * @sk_stamp - time stamp of last packet received
166 * @sk_socket - Identd and reporting IO signals
167 * @sk_user_data - RPC layer private data
168 * @sk_owner - module that owns this socket
169 * @sk_sndmsg_page - cached page for sendmsg
170 * @sk_sndmsg_off - cached offset for sendmsg
171 * @sk_send_head - front of stuff to transmit
172 * @sk_write_pending - a write to stream socket waits to start
173 * @sk_queue_shrunk - write queue has been shrunk recently
174 * @sk_state_change - callback to indicate change in the state of the sock
175 * @sk_data_ready - callback to indicate there is data to be processed
176 * @sk_write_space - callback to indicate there is bf sending space available
177 * @sk_error_report - callback to indicate errors (e.g. %MSG_ERRQUEUE)
178 * @sk_backlog_rcv - callback to process the backlog
179 * @sk_destruct - called at sock freeing time, i.e. when all refcnt == 0
183 * Now struct tcp_tw_bucket also uses sock_common, so please just
184 * don't add nothing before this first member (__sk_common) --acme
186 struct sock_common __sk_common;
187 #define sk_family __sk_common.skc_family
188 #define sk_state __sk_common.skc_state
189 #define sk_reuse __sk_common.skc_reuse
190 #define sk_bound_dev_if __sk_common.skc_bound_dev_if
191 #define sk_node __sk_common.skc_node
192 #define sk_bind_node __sk_common.skc_bind_node
193 #define sk_refcnt __sk_common.skc_refcnt
194 volatile unsigned char sk_zapped;
195 unsigned char sk_shutdown;
196 unsigned char sk_use_write_queue;
197 unsigned char sk_userlocks;
198 socket_lock_t sk_lock;
200 wait_queue_head_t *sk_sleep;
201 struct dst_entry *sk_dst_cache;
202 rwlock_t sk_dst_lock;
203 struct xfrm_policy *sk_policy[2];
204 atomic_t sk_rmem_alloc;
205 struct sk_buff_head sk_receive_queue;
206 atomic_t sk_wmem_alloc;
207 struct sk_buff_head sk_write_queue;
208 atomic_t sk_omem_alloc;
210 int sk_forward_alloc;
211 unsigned int sk_allocation;
213 unsigned long sk_flags;
215 unsigned char sk_debug;
216 unsigned char sk_rcvtstamp;
217 unsigned char sk_no_largesend;
219 unsigned long sk_lingertime;
222 * The backlog queue is special, it is always used with
223 * the per-socket spinlock held and requires low latency
224 * access. Therefore we special case it's implementation.
227 struct sk_buff *head;
228 struct sk_buff *tail;
230 rwlock_t sk_callback_lock;
231 struct sk_buff_head sk_error_queue;
232 struct proto *sk_prot;
235 unsigned short sk_ack_backlog;
236 unsigned short sk_max_ack_backlog;
238 unsigned short sk_type;
239 unsigned char sk_localroute;
240 unsigned char sk_protocol;
241 struct ucred sk_peercred;
245 struct sk_filter *sk_filter;
247 kmem_cache_t *sk_slab;
248 struct timer_list sk_timer;
249 struct timeval sk_stamp;
250 struct socket *sk_socket;
252 void *sk_ns; // For use by CKRM
253 struct module *sk_owner;
254 struct page *sk_sndmsg_page;
256 struct sk_buff *sk_send_head;
257 int sk_write_pending;
259 __u8 sk_queue_shrunk;
260 /* three bytes hole, try to pack */
261 void (*sk_state_change)(struct sock *sk);
262 void (*sk_data_ready)(struct sock *sk, int bytes);
263 void (*sk_write_space)(struct sock *sk);
264 void (*sk_error_report)(struct sock *sk);
265 int (*sk_backlog_rcv)(struct sock *sk,
266 struct sk_buff *skb);
267 void (*sk_destruct)(struct sock *sk);
271 * Hashed lists helper routines
273 static inline struct sock *__sk_head(struct hlist_head *head)
275 return hlist_entry(head->first, struct sock, sk_node);
278 static inline struct sock *sk_head(struct hlist_head *head)
280 return hlist_empty(head) ? NULL : __sk_head(head);
283 static inline struct sock *sk_next(struct sock *sk)
285 return sk->sk_node.next ?
286 hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
289 static inline int sk_unhashed(struct sock *sk)
291 return hlist_unhashed(&sk->sk_node);
294 static inline int sk_hashed(struct sock *sk)
296 return sk->sk_node.pprev != NULL;
299 static __inline__ void sk_node_init(struct hlist_node *node)
304 static __inline__ void __sk_del_node(struct sock *sk)
306 __hlist_del(&sk->sk_node);
309 static __inline__ int __sk_del_node_init(struct sock *sk)
313 sk_node_init(&sk->sk_node);
319 /* Grab socket reference count. This operation is valid only
320 when sk is ALREADY grabbed f.e. it is found in hash table
321 or a list and the lookup is made under lock preventing hash table
325 static inline void sock_hold(struct sock *sk)
327 atomic_inc(&sk->sk_refcnt);
330 /* Ungrab socket in the context, which assumes that socket refcnt
331 cannot hit zero, f.e. it is true in context of any socketcall.
333 static inline void __sock_put(struct sock *sk)
335 atomic_dec(&sk->sk_refcnt);
338 static __inline__ int sk_del_node_init(struct sock *sk)
340 int rc = __sk_del_node_init(sk);
343 /* paranoid for a while -acme */
344 WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
350 static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list)
352 hlist_add_head(&sk->sk_node, list);
355 static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list)
358 __sk_add_node(sk, list);
361 static __inline__ void __sk_del_bind_node(struct sock *sk)
363 __hlist_del(&sk->sk_bind_node);
366 static __inline__ void sk_add_bind_node(struct sock *sk,
367 struct hlist_head *list)
369 hlist_add_head(&sk->sk_bind_node, list);
372 #define sk_for_each(__sk, node, list) \
373 hlist_for_each_entry(__sk, node, list, sk_node)
374 #define sk_for_each_from(__sk, node) \
375 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
376 hlist_for_each_entry_from(__sk, node, sk_node)
377 #define sk_for_each_continue(__sk, node) \
378 if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
379 hlist_for_each_entry_continue(__sk, node, sk_node)
380 #define sk_for_each_safe(__sk, node, tmp, list) \
381 hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
382 #define sk_for_each_bound(__sk, node, list) \
383 hlist_for_each_entry(__sk, node, list, sk_bind_node)
397 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
399 __set_bit(flag, &sk->sk_flags);
402 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
404 __clear_bit(flag, &sk->sk_flags);
407 static inline int sock_flag(struct sock *sk, enum sock_flags flag)
409 return test_bit(flag, &sk->sk_flags);
412 #ifndef CONFIG_ACCEPT_QUEUES
413 static inline void sk_acceptq_removed(struct sock *sk)
415 sk->sk_ack_backlog--;
418 static inline void sk_acceptq_added(struct sock *sk)
420 sk->sk_ack_backlog++;
423 static inline int sk_acceptq_is_full(struct sock *sk)
425 return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
430 * Compute minimal free write space needed to queue new packets.
432 static inline int sk_stream_min_wspace(struct sock *sk)
434 return sk->sk_wmem_queued / 2;
437 static inline int sk_stream_wspace(struct sock *sk)
439 return sk->sk_sndbuf - sk->sk_wmem_queued;
442 extern void sk_stream_write_space(struct sock *sk);
444 static inline int sk_stream_memory_free(struct sock *sk)
446 return sk->sk_wmem_queued < sk->sk_sndbuf;
449 extern void sk_stream_rfree(struct sk_buff *skb);
451 static inline void sk_stream_set_owner_r(struct sk_buff *skb, struct sock *sk)
454 skb->destructor = sk_stream_rfree;
455 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
456 sk->sk_forward_alloc -= skb->truesize;
459 static inline void sk_stream_free_skb(struct sock *sk, struct sk_buff *skb)
461 sk->sk_queue_shrunk = 1;
462 sk->sk_wmem_queued -= skb->truesize;
463 sk->sk_forward_alloc += skb->truesize;
467 /* The per-socket spinlock must be held here. */
468 #define sk_add_backlog(__sk, __skb) \
469 do { if (!(__sk)->sk_backlog.tail) { \
470 (__sk)->sk_backlog.head = \
471 (__sk)->sk_backlog.tail = (__skb); \
473 ((__sk)->sk_backlog.tail)->next = (__skb); \
474 (__sk)->sk_backlog.tail = (__skb); \
476 (__skb)->next = NULL; \
479 #define sk_wait_event(__sk, __timeo, __condition) \
481 release_sock(__sk); \
484 *(__timeo) = schedule_timeout(*(__timeo)); \
491 extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
492 extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
493 extern void sk_stream_wait_close(struct sock *sk, long timeo_p);
494 extern int sk_stream_error(struct sock *sk, int flags, int err);
495 extern void sk_stream_kill_queues(struct sock *sk);
497 extern int sk_wait_data(struct sock *sk, long *timeo);
499 /* Networking protocol blocks we attach to sockets.
500 * socket layer -> transport layer interface
501 * transport -> network interface is defined by struct inet_proto
504 void (*close)(struct sock *sk,
506 int (*connect)(struct sock *sk,
507 struct sockaddr *uaddr,
509 int (*disconnect)(struct sock *sk, int flags);
511 struct sock * (*accept) (struct sock *sk, int flags, int *err);
513 int (*ioctl)(struct sock *sk, int cmd,
515 int (*init)(struct sock *sk);
516 int (*destroy)(struct sock *sk);
517 void (*shutdown)(struct sock *sk, int how);
518 int (*setsockopt)(struct sock *sk, int level,
519 int optname, char __user *optval,
521 int (*getsockopt)(struct sock *sk, int level,
522 int optname, char __user *optval,
524 int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
525 struct msghdr *msg, size_t len);
526 int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
528 size_t len, int noblock, int flags,
530 int (*sendpage)(struct sock *sk, struct page *page,
531 int offset, size_t size, int flags);
532 int (*bind)(struct sock *sk,
533 struct sockaddr *uaddr, int addr_len);
535 int (*backlog_rcv) (struct sock *sk,
536 struct sk_buff *skb);
538 /* Keeping track of sk's, looking them up, and port selection methods. */
539 void (*hash)(struct sock *sk);
540 void (*unhash)(struct sock *sk);
541 int (*get_port)(struct sock *sk, unsigned short snum);
543 /* Memory pressure */
544 void (*enter_memory_pressure)(void);
545 atomic_t *memory_allocated; /* Current allocated memory. */
546 atomic_t *sockets_allocated; /* Current number of sockets. */
548 * Pressure flag: try to collapse.
549 * Technical note: it is used by multiple contexts non atomically.
550 * All the sk_stream_mem_schedule() is of this nature: accounting
551 * is strict, actions are advisory and have some latency.
553 int *memory_pressure;
562 struct module *owner;
568 u8 __pad[SMP_CACHE_BYTES - sizeof(int)];
572 extern int sk_alloc_slab(struct proto *prot, char *name);
573 extern void sk_free_slab(struct proto *prot);
575 static inline void sk_alloc_slab_error(struct proto *proto)
577 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n", proto->name);
580 static __inline__ void sk_set_owner(struct sock *sk, struct module *owner)
583 * One should use sk_set_owner just once, after struct sock creation,
584 * be it shortly after sk_alloc or after a function that returns a new
585 * struct sock (and that down the call chain called sk_alloc), e.g. the
586 * IPv4 and IPv6 modules share tcp_create_openreq_child, so if
587 * tcp_create_openreq_child called sk_set_owner IPv6 would have to
588 * change the ownership of this struct sock, with one not needed
589 * transient sk_set_owner call.
591 BUG_ON(sk->sk_owner != NULL);
593 sk->sk_owner = owner;
597 /* Called with local bh disabled */
598 static __inline__ void sock_prot_inc_use(struct proto *prot)
600 prot->stats[smp_processor_id()].inuse++;
603 static __inline__ void sock_prot_dec_use(struct proto *prot)
605 prot->stats[smp_processor_id()].inuse--;
608 /* About 10 seconds */
609 #define SOCK_DESTROY_TIME (10*HZ)
611 /* Sockets 0-1023 can't be bound to unless you are superuser */
612 #define PROT_SOCK 1024
614 #define SHUTDOWN_MASK 3
615 #define RCV_SHUTDOWN 1
616 #define SEND_SHUTDOWN 2
618 #define SOCK_SNDBUF_LOCK 1
619 #define SOCK_RCVBUF_LOCK 2
620 #define SOCK_BINDADDR_LOCK 4
621 #define SOCK_BINDPORT_LOCK 8
623 /* sock_iocb: used to kick off async processing of socket ios */
625 struct list_head list;
631 struct scm_cookie *scm;
632 struct msghdr *msg, async_msg;
633 struct iovec async_iov;
637 static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
639 return (struct sock_iocb *)iocb->private;
642 static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
647 struct socket_alloc {
648 struct socket socket;
649 struct inode vfs_inode;
652 static inline struct socket *SOCKET_I(struct inode *inode)
654 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
657 static inline struct inode *SOCK_INODE(struct socket *socket)
659 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
662 extern void __sk_stream_mem_reclaim(struct sock *sk);
663 extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
665 #define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
667 static inline int sk_stream_pages(int amt)
669 return (amt + SK_STREAM_MEM_QUANTUM - 1) / SK_STREAM_MEM_QUANTUM;
672 static inline void sk_stream_mem_reclaim(struct sock *sk)
674 if (sk->sk_forward_alloc >= SK_STREAM_MEM_QUANTUM)
675 __sk_stream_mem_reclaim(sk);
678 static inline void sk_stream_writequeue_purge(struct sock *sk)
682 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
683 sk_stream_free_skb(sk, skb);
684 sk_stream_mem_reclaim(sk);
687 static inline int sk_stream_rmem_schedule(struct sock *sk, struct sk_buff *skb)
689 return (int)skb->truesize <= sk->sk_forward_alloc ||
690 sk_stream_mem_schedule(sk, skb->truesize, 1);
693 /* Used by processes to "lock" a socket state, so that
694 * interrupts and bottom half handlers won't change it
695 * from under us. It essentially blocks any incoming
696 * packets, so that we won't get any new data or any
697 * packets that change the state of the socket.
699 * While locked, BH processing will add new packets to
700 * the backlog queue. This queue is processed by the
701 * owner of the socket lock right before it is released.
703 * Since ~2.3.5 it is also exclusive sleep lock serializing
704 * accesses from user process context.
706 #define sock_owned_by_user(sk) ((sk)->sk_lock.owner)
708 extern void FASTCALL(lock_sock(struct sock *sk));
709 extern void FASTCALL(release_sock(struct sock *sk));
711 /* BH context may only use the following locking interface. */
712 #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
713 #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
715 extern struct sock * sk_alloc(int family, int priority, int zero_it,
717 extern void sk_free(struct sock *sk);
719 extern struct sk_buff *sock_wmalloc(struct sock *sk,
720 unsigned long size, int force,
722 extern struct sk_buff *sock_rmalloc(struct sock *sk,
723 unsigned long size, int force,
725 extern void sock_wfree(struct sk_buff *skb);
726 extern void sock_rfree(struct sk_buff *skb);
728 extern int sock_setsockopt(struct socket *sock, int level,
729 int op, char __user *optval,
732 extern int sock_getsockopt(struct socket *sock, int level,
733 int op, char __user *optval,
735 extern struct sk_buff *sock_alloc_send_skb(struct sock *sk,
739 extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
740 unsigned long header_len,
741 unsigned long data_len,
744 extern void *sock_kmalloc(struct sock *sk, int size, int priority);
745 extern void sock_kfree_s(struct sock *sk, void *mem, int size);
746 extern void sk_send_sigurg(struct sock *sk);
749 * Functions to fill in entries in struct proto_ops when a protocol
750 * does not implement a particular function.
752 extern int sock_no_bind(struct socket *,
753 struct sockaddr *, int);
754 extern int sock_no_connect(struct socket *,
755 struct sockaddr *, int, int);
756 extern int sock_no_socketpair(struct socket *,
758 extern int sock_no_accept(struct socket *,
759 struct socket *, int);
760 extern int sock_no_getname(struct socket *,
761 struct sockaddr *, int *, int);
762 extern unsigned int sock_no_poll(struct file *, struct socket *,
763 struct poll_table_struct *);
764 extern int sock_no_ioctl(struct socket *, unsigned int,
766 extern int sock_no_listen(struct socket *, int);
767 extern int sock_no_shutdown(struct socket *, int);
768 extern int sock_no_getsockopt(struct socket *, int , int,
769 char __user *, int __user *);
770 extern int sock_no_setsockopt(struct socket *, int, int,
772 extern int sock_no_sendmsg(struct kiocb *, struct socket *,
773 struct msghdr *, size_t);
774 extern int sock_no_recvmsg(struct kiocb *, struct socket *,
775 struct msghdr *, size_t, int);
776 extern int sock_no_mmap(struct file *file,
778 struct vm_area_struct *vma);
779 extern ssize_t sock_no_sendpage(struct socket *sock,
781 int offset, size_t size,
785 * Functions to fill in entries in struct proto_ops when a protocol
786 * uses the inet style.
788 extern int sock_common_getsockopt(struct socket *sock, int level, int optname,
789 char __user *optval, int __user *optlen);
790 extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
791 struct msghdr *msg, size_t size, int flags);
792 extern int sock_common_setsockopt(struct socket *sock, int level, int optname,
793 char __user *optval, int optlen);
795 extern void sk_common_release(struct sock *sk);
798 * Default socket callbacks and setup code
801 extern void sock_def_destruct(struct sock *);
803 /* Initialise core socket variables */
804 extern void sock_init_data(struct socket *sock, struct sock *sk);
807 * sk_filter - run a packet through a socket filter
808 * @sk: sock associated with &sk_buff
809 * @skb: buffer to filter
810 * @needlock: set to 1 if the sock is not locked by caller.
812 * Run the filter code and then cut skb->data to correct size returned by
813 * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller
814 * than pkt_len we keep whole skb->data. This is the socket level
815 * wrapper to sk_run_filter. It returns 0 if the packet should
816 * be accepted or -EPERM if the packet should be tossed.
820 static inline int sk_filter(struct sock *sk, struct sk_buff *skb, int needlock)
824 err = security_sock_rcv_skb(sk, skb);
829 struct sk_filter *filter;
834 filter = sk->sk_filter;
836 int pkt_len = sk_run_filter(skb, filter->insns,
841 skb_trim(skb, pkt_len);
851 * sk_filter_release: Release a socket filter
853 * @fp: filter to remove
855 * Remove a filter from a socket and release its resources.
858 static inline void sk_filter_release(struct sock *sk, struct sk_filter *fp)
860 unsigned int size = sk_filter_len(fp);
862 atomic_sub(size, &sk->sk_omem_alloc);
864 if (atomic_dec_and_test(&fp->refcnt))
868 static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
870 atomic_inc(&fp->refcnt);
871 atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
875 * Socket reference counting postulates.
877 * * Each user of socket SHOULD hold a reference count.
878 * * Each access point to socket (an hash table bucket, reference from a list,
879 * running timer, skb in flight MUST hold a reference count.
880 * * When reference count hits 0, it means it will never increase back.
881 * * When reference count hits 0, it means that no references from
882 * outside exist to this socket and current process on current CPU
883 * is last user and may/should destroy this socket.
884 * * sk_free is called from any context: process, BH, IRQ. When
885 * it is called, socket has no references from outside -> sk_free
886 * may release descendant resources allocated by the socket, but
887 * to the time when it is called, socket is NOT referenced by any
888 * hash tables, lists etc.
889 * * Packets, delivered from outside (from network or from another process)
890 * and enqueued on receive/error queues SHOULD NOT grab reference count,
891 * when they sit in queue. Otherwise, packets will leak to hole, when
892 * socket is looked up by one cpu and unhasing is made by another CPU.
893 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
894 * (leak to backlog). Packet socket does all the processing inside
895 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
896 * use separate SMP lock, so that they are prone too.
899 /* Ungrab socket and destroy it, if it was the last reference. */
900 static inline void sock_put(struct sock *sk)
902 if (atomic_dec_and_test(&sk->sk_refcnt))
906 /* Detach socket from process context.
907 * Announce socket dead, detach it from wait queue and inode.
908 * Note that parent inode held reference count on this struct sock,
909 * we do not release it in this function, because protocol
910 * probably wants some additional cleanups or even continuing
911 * to work with this socket (TCP).
913 static inline void sock_orphan(struct sock *sk)
915 write_lock_bh(&sk->sk_callback_lock);
916 sock_set_flag(sk, SOCK_DEAD);
917 sk->sk_socket = NULL;
919 write_unlock_bh(&sk->sk_callback_lock);
922 static inline void sock_graft(struct sock *sk, struct socket *parent)
924 write_lock_bh(&sk->sk_callback_lock);
925 sk->sk_sleep = &parent->wait;
927 sk->sk_socket = parent;
928 write_unlock_bh(&sk->sk_callback_lock);
931 extern int sock_i_uid(struct sock *sk);
932 extern unsigned long sock_i_ino(struct sock *sk);
934 static inline struct dst_entry *
935 __sk_dst_get(struct sock *sk)
937 return sk->sk_dst_cache;
940 static inline struct dst_entry *
941 sk_dst_get(struct sock *sk)
943 struct dst_entry *dst;
945 read_lock(&sk->sk_dst_lock);
946 dst = sk->sk_dst_cache;
949 read_unlock(&sk->sk_dst_lock);
954 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
956 struct dst_entry *old_dst;
958 old_dst = sk->sk_dst_cache;
959 sk->sk_dst_cache = dst;
960 dst_release(old_dst);
964 sk_dst_set(struct sock *sk, struct dst_entry *dst)
966 write_lock(&sk->sk_dst_lock);
967 __sk_dst_set(sk, dst);
968 write_unlock(&sk->sk_dst_lock);
972 __sk_dst_reset(struct sock *sk)
974 struct dst_entry *old_dst;
976 old_dst = sk->sk_dst_cache;
977 sk->sk_dst_cache = NULL;
978 dst_release(old_dst);
982 sk_dst_reset(struct sock *sk)
984 write_lock(&sk->sk_dst_lock);
986 write_unlock(&sk->sk_dst_lock);
989 static inline struct dst_entry *
990 __sk_dst_check(struct sock *sk, u32 cookie)
992 struct dst_entry *dst = sk->sk_dst_cache;
994 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
995 sk->sk_dst_cache = NULL;
1002 static inline struct dst_entry *
1003 sk_dst_check(struct sock *sk, u32 cookie)
1005 struct dst_entry *dst = sk_dst_get(sk);
1007 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
1015 static inline void sk_charge_skb(struct sock *sk, struct sk_buff *skb)
1017 sk->sk_wmem_queued += skb->truesize;
1018 sk->sk_forward_alloc -= skb->truesize;
1021 static inline int skb_copy_to_page(struct sock *sk, char __user *from,
1022 struct sk_buff *skb, struct page *page,
1025 if (skb->ip_summed == CHECKSUM_NONE) {
1027 unsigned int csum = csum_and_copy_from_user(from,
1028 page_address(page) + off,
1032 skb->csum = csum_block_add(skb->csum, csum, skb->len);
1033 } else if (copy_from_user(page_address(page) + off, from, copy))
1037 skb->data_len += copy;
1038 skb->truesize += copy;
1039 sk->sk_wmem_queued += copy;
1040 sk->sk_forward_alloc -= copy;
1045 * Queue a received datagram if it will fit. Stream and sequenced
1046 * protocols can't normally use this as they need to fit buffers in
1047 * and play with them.
1049 * Inlined as it's very short and called for pretty much every
1050 * packet ever received.
1053 static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1057 skb->destructor = sock_wfree;
1058 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1061 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1064 skb->destructor = sock_rfree;
1065 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1068 extern void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1069 unsigned long expires);
1071 extern void sk_stop_timer(struct sock *sk, struct timer_list* timer);
1073 static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
1078 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1079 number of warnings when compiling with -W --ANK
1081 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1082 (unsigned)sk->sk_rcvbuf) {
1087 /* It would be deadlock, if sock_queue_rcv_skb is used
1088 with socket lock! We assume that users of this
1089 function are lock free.
1091 err = sk_filter(sk, skb, 1);
1096 skb_set_owner_r(skb, sk);
1098 /* Cache the SKB length before we tack it onto the receive
1099 * queue. Once it is added it no longer belongs to us and
1100 * may be freed by other threads of control pulling packets
1105 skb_queue_tail(&sk->sk_receive_queue, skb);
1107 if (!sock_flag(sk, SOCK_DEAD))
1108 sk->sk_data_ready(sk, skb_len);
1113 static inline int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb)
1115 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
1116 number of warnings when compiling with -W --ANK
1118 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
1119 (unsigned)sk->sk_rcvbuf)
1121 skb_set_owner_r(skb, sk);
1122 skb_queue_tail(&sk->sk_error_queue, skb);
1123 if (!sock_flag(sk, SOCK_DEAD))
1124 sk->sk_data_ready(sk, skb->len);
1129 * Recover an error report and clear atomically
1132 static inline int sock_error(struct sock *sk)
1134 int err = xchg(&sk->sk_err, 0);
1138 static inline unsigned long sock_wspace(struct sock *sk)
1142 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1143 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1150 static inline void sk_wake_async(struct sock *sk, int how, int band)
1152 if (sk->sk_socket && sk->sk_socket->fasync_list)
1153 sock_wake_async(sk->sk_socket, how, band);
1156 #define SOCK_MIN_SNDBUF 2048
1157 #define SOCK_MIN_RCVBUF 256
1159 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1161 if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1162 sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued / 2);
1163 sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1167 static inline struct sk_buff *sk_stream_alloc_pskb(struct sock *sk,
1168 int size, int mem, int gfp)
1170 struct sk_buff *skb = alloc_skb(size + sk->sk_prot->max_header, gfp);
1173 skb->truesize += mem;
1174 if (sk->sk_forward_alloc >= (int)skb->truesize ||
1175 sk_stream_mem_schedule(sk, skb->truesize, 0)) {
1176 skb_reserve(skb, sk->sk_prot->max_header);
1181 sk->sk_prot->enter_memory_pressure();
1182 sk_stream_moderate_sndbuf(sk);
1187 static inline struct sk_buff *sk_stream_alloc_skb(struct sock *sk,
1190 return sk_stream_alloc_pskb(sk, size, 0, gfp);
1193 static inline struct page *sk_stream_alloc_page(struct sock *sk)
1195 struct page *page = NULL;
1197 if (sk->sk_forward_alloc >= (int)PAGE_SIZE ||
1198 sk_stream_mem_schedule(sk, PAGE_SIZE, 0))
1199 page = alloc_pages(sk->sk_allocation, 0);
1201 sk->sk_prot->enter_memory_pressure();
1202 sk_stream_moderate_sndbuf(sk);
1207 #define sk_stream_for_retrans_queue(skb, sk) \
1208 for (skb = (sk)->sk_write_queue.next; \
1209 (skb != (sk)->sk_send_head) && \
1210 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1214 * Default write policy as shown to user space via poll/select/SIGIO
1216 static inline int sock_writeable(const struct sock *sk)
1218 return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf / 2);
1221 static inline int gfp_any(void)
1223 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1226 static inline long sock_rcvtimeo(const struct sock *sk, int noblock)
1228 return noblock ? 0 : sk->sk_rcvtimeo;
1231 static inline long sock_sndtimeo(const struct sock *sk, int noblock)
1233 return noblock ? 0 : sk->sk_sndtimeo;
1236 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1238 return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
1241 /* Alas, with timeout socket operations are not restartable.
1242 * Compare this to poll().
1244 static inline int sock_intr_errno(long timeo)
1246 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1249 static __inline__ void
1250 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
1252 struct timeval *stamp = &skb->stamp;
1253 if (sk->sk_rcvtstamp) {
1254 /* Race occurred between timestamp enabling and packet
1255 receiving. Fill in the current time for now. */
1256 if (stamp->tv_sec == 0)
1257 do_gettimeofday(stamp);
1258 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP, sizeof(struct timeval),
1261 sk->sk_stamp = *stamp;
1265 * sk_eat_skb - Release a skb if it is no longer needed
1266 * @sk - socket to eat this skb from
1267 * @skb - socket buffer to eat
1269 * This routine must be called with interrupts disabled or with the socket
1270 * locked so that the sk_buff queue operation is ok.
1272 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1274 __skb_unlink(skb, &sk->sk_receive_queue);
1278 extern void sock_enable_timestamp(struct sock *sk);
1279 extern int sock_get_timestamp(struct sock *, struct timeval __user *);
1282 * Enable debug/info messages
1286 #define NETDEBUG(x) do { } while (0)
1287 #define LIMIT_NETDEBUG(x) do {} while(0)
1289 #define NETDEBUG(x) do { x; } while (0)
1290 #define LIMIT_NETDEBUG(x) do { if (net_ratelimit()) { x; } } while(0)
1294 * Macros for sleeping on a socket. Use them like this:
1296 * SOCK_SLEEP_PRE(sk)
1299 * SOCK_SLEEP_POST(sk)
1301 * N.B. These are now obsolete and were, afaik, only ever used in DECnet
1302 * and when the last use of them in DECnet has gone, I'm intending to
1306 #define SOCK_SLEEP_PRE(sk) { struct task_struct *tsk = current; \
1307 DECLARE_WAITQUEUE(wait, tsk); \
1308 tsk->state = TASK_INTERRUPTIBLE; \
1309 add_wait_queue((sk)->sk_sleep, &wait); \
1312 #define SOCK_SLEEP_POST(sk) tsk->state = TASK_RUNNING; \
1313 remove_wait_queue((sk)->sk_sleep, &wait); \
1317 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
1320 sock_set_flag(sk, bit);
1322 sock_reset_flag(sk, bit);
1325 extern __u32 sysctl_wmem_max;
1326 extern __u32 sysctl_rmem_max;
1329 int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg);
1331 static inline int siocdevprivate_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
1337 #endif /* _SOCK_H */